AI Risks "Hypernormal" Science

The article argues that scaling AI systems won't automatically produce scientific paradigm shifts. While AI excels at mapping and navigating existing knowledge—like AlphaFold predicting protein structures—it risks reinforcing 'normal science' by finding more of the same kind of information rather than fundamentally new frameworks. True scientific breakthroughs, like Maxwell's unification of electricity and magnetism, require reimagining the underlying schematic, not just adding detail to existing maps.